At 11.54 +0200 2001-05-21, Hans Aberg wrote:
>So the best one could hope for, I think, is to build a layer above those
>glyphs, say an empty set command that can expand to \emptyset or
>\varnothing, a set membership command that can expand to either of \in or
>the two epsilon variations, but there can be no restrictions on the glyphs
>themselves in the sense that the use of one of them prohibits the other.

My suggestion is rather that the current set of math symbol commands should
be made into this layer (hence some commands which now produce different
results would then by default produce identical results). It should still
be possible to access to any glyph available in the fonts, but the author
who wishes to do so must be prepared to first define a new command for
accessing the glyph via some mixed-case command (cf. how one uses
\DeclareTextSymbol for text today). There could also be a compability
package which makes all commands that produce distinct results today
continue to do this. Finally, for the more common task of selecting which
glyphic variant of a character to use, there should be a simpler interface
than giving an explicit code point, perhaps something like

\chooseglyph{\leq}{slanted}

to make \leq produce what \leqslant does today. (If that glyph is not
available, a warning like

Warning: Glyph variant slanted' of command \leq not defined.
Using std' variant instead.

should be issued. \leqq and \eqslantless could also be considered glyph
variants of \leq, unless they actually have distinct meanings somewhere.)

Lars Hellström